Real Agriculture currently has a nice interview with agronomist Ross McKenzie on the historic impact of agriculture on soil health. In this video, Ross describes the important role of organic matter (e.g. old crop debris) in soil health. These decomposing plant bits act as a natural glue, clumping the soil into granules that improve soil structure by creating pores.
One of my all-time favorite scientific figures from one of my favorite books, Water Relations of Plants and Soils, illustrates the concept of soil structure. Capillary pores are microscopic pits in soil particles that hold water (and nutrients) more tightly than gravity can pull them away. Non-capillary pores are larger (e.g. spaces left after roots decay) and are critical for proper aeration and water movement. This figure shows the natural pore structure of forest soil versus heavily-compacted farm soil. The lack of pore space in the farm soil prevents the absorption (and retention) of rain, leading to erosion and dry, nutrient-poor soils. As this figure demonstrates, soil structure can be compacted as deep as 2 feet by agricultural practices (e.g. driving heavy equipment on wet fields).
Soil texture (the distribution of soil particle sizes) is also critical. Extremely tiny particles (clay) hold the most water, but bind it too tightly for plants to completely utilize. Extremely large particles (sand) hold very little water, but bind it weakly. Excellent garden soil has a balanced soil structure and texture, low acidity and is black with organic matter.
Tillage was easily one of the most important inventions of early human history. This process, whereby the top few inches of soil are lifted up and mixed, aerates and loosens the topsoil, buries crop residue and kills pests, pathogens and weeds. Unfortunately, tillage also hastens the breakdown of organic matter, compacts the deeper layers of soil and leads to rampant erosion. It also has subtle ecological effects. One of my old professors once told me about a study of his that found that plows didn't kill many of the countless, microscopic, plant-parasitic nematode worms that inhabit farm soils, but DID kill their larger, rarer predator relatives.
Tillage worked pretty well in Europe, but when applied writ large in North America, it created the rampant erosion that led to the Dust Bowl. This picture* shows typical root growth produced by annual crops versus their wild, perennial relatives. These massive perennial root masses formed the thick sod that supported and protected North American prairie soils prior to European colonization.
Modern conservation tillage (aka no-till) is a tactic that attempts to build ecologically-functional soils while protecting the crop from weeds, pests and pathogens by other methods (e.g. planting cold-tolerant varieties very early in the year, planting varieties with transgenic (GM) glyphosate resistance and applying chemical pesticides).
Theoretically, it would be extremely sustainable to develop perennial versions of our grain and legume crops with which to establish vast semi-natural "prairies," but this goal is much more difficult than it may seem and success is likely a long way off. Relatively few public sector labs are working on this goal, although the non-profit, The Land Institute, claims to be close to releasing varieties. Private seed companies can't be expected to contribute to this goal as it would appear to be an irresponsible business plan (a good reason to increase public funding of agricultural science!).
*Picture from "Prospects for Developing Perennial Grain Crops" BioScience August 2006 Vol. 56 No. 8